Path: utzoo!news-server.csri.toronto.edu!rutgers!usc!nic.csu.net!csun!kithrup!sef
From: sef@kithrup.COM (Sean Eric Fagan)
Newsgroups: comp.arch
Subject: Re: Novice question:  measuring speed
Message-ID: <1991Mar13.102841.7854@kithrup.COM>
Date: 13 Mar 91 10:28:41 GMT
References: <645@ssdc?>
Organization: Kithrup Enterprises, Ltd.
Lines: 51

In article <645@ssdc?> jbasara@ssdc (jim basara) writes:
>Could someone please
>provide me with a descriptive explaination of why MIP ratings are not a 
>good way of comparing processing speed for RISC machines as opposed to MFLOPS?

MFLOPS aren't that great either, but there *is* a difference.

Let's imagine a machine which can do an instruction fetch and decode in once
cycle, and execute the instruction in another.  Add one cycle for every
"operation" it does (such as load from memory, store to memory, etc.).  An
instruction such as

	add	$4, [0x100]	; encoded as 0x01040100

would take 4 cycles to fetch (assuming 8-bit bus, of course), one cycle to
load from memory location 0x100, one cycle to add 4 to it, and one cycle to
write the new value to location 0x100.  A total of 7 cycles, no?  If the
machine ran at 10MHz, and we only consider that instruction, then we're
talking about a slightly-larger-than-1 MIPS machinne, no?  Now let's
consider another instruction the processor has:  NOP.  NOP is encoded as
0xea.  NOP takes one cycle to fetch and decode, and 0 cycles to run.  A
total of 1 cycle.  Therefore, based on that instruction, the machine is a
10MIPS machine at 10MHz.

Both measurements are, technically, correct.

Now, as to why MFLOPS are different:  because MFLOPS doesn't mean "Millions
of FLoating-point instructions Per Second," it means "Millions of FLoating
point Operations Per Second."  A nop does (by definition) no operation on a
floating-point value (although I've seen 'FNOP' instructions 8-)), so would
not be counted.  As a result, the measurements for MFLOPS tend to be a
*little* more accurate than those for MIPS.

All in all, though, go for SPECmarks.  Much better.

>I would also like to know why companies are using MIP ratings for those machine
>s if they are not accurate.  

Sales.

>Is it possible that the MIP ratings can be 
>manipulated so that a processor with a higher MIP rating may actually be
>considerably slower than one with a lower MIP rating???  

I think I answered that with my demonstration above.

-- 
Sean Eric Fagan  | "I made the universe, but please don't blame me for it;
sef@kithrup.COM  |  I had a bellyache at the time."
-----------------+           -- The Turtle (Stephen King, _It_)
Any opinions expressed are my own, and generally unpopular with others.